Define behaviour when resuming an interupted run

[andrewphilipsmith]

As a:
user who is processing large volumes of data
I want:
alto2txt to be able to resume elegantly when run on a partially complete previous run, including cases where the previous run had crashed.
so that:
I can resume it quickly without needing to manually wade through lengthy log files, nor needing to find partial, incomplete output files.

Notes:

• We will need to consider scenarios where (i) alto2txt is run in serial and (ii) alto2txt is run in parallel.
• The current behaviour in these scenarios is unknown.

This idea came out of a meeting with @fedenanni and @thobson88. Our notes are here https://hackmd.io/KeOzeaMYTOiF37pGUjUl7A